To predict the power of photovoltaic power generation, we analyzed the factors of photovoltaic power generation and proposed a method of predicting photovoltaic power generation power based on the curve matching of solar radiation. By decomposing the historical data, searching the solar radiation curve, using the BP neural network, this method could predict photovoltaic power generation power within the next three hours and have a good performance. Experimental results show that this model can have a small forecasting error, which means that it’s important in electric network management.